Hi David, On Sat, Feb 04, 2023 at 01:45:15PM +0100, David Virag wrote: > On Sat, 2023-02-04 at 12:43 +0100, David Virag wrote: > > On Fri, 2023-02-03 at 21:39 +0100, Henrik Grimler wrote: > > > They are no longer needed after commit a13e8ef6008d ("mmc: dw_mmc: > > > exynos: use common_caps"). > > > > > > Signed-off-by: Henrik Grimler <henrik@xxxxxxxxxx> > > > --- > > > arch/arm64/boot/dts/exynos/exynos5433-tm2-common.dtsi | 2 -- > > > arch/arm64/boot/dts/exynos/exynos7-espresso.dts | 2 -- > > > 2 files changed, 4 deletions(-) > [...] > > In dw_mmc-exynos.c, there's a caps array specified like this: > > > > /* Common capabilities of Exynos4/Exynos5 SoC */ > > static unsigned long exynos_dwmmc_caps[4] = { > > MMC_CAP_1_8V_DDR | MMC_CAP_8_BIT_DATA, > > 0, > > 0, > > 0, > > }; > > > > As I understand these capabilities are added to the mmc controllers > > based on mshc alias id. Shouldn't these capabilities be moved to > > device-tree before removing these aliases? This also applies to the > > 32bit arm patch. If I understand correctly, removing these aliases > > without adding the capabilities to dt removes the capability > > "MMC_CAP_1_8V_DDR" and "MMC_CAP_8_BIT_DATA" from mshc_0/mmc_0. Thanks for pointing this out (I should have seen it already), will send a new version to update device trees and remove the need for those mshc alias based capabilities. > Actually, it defaults to 0, so these capabilities will be added to all > mmc nodes, not just those with mshc0 alias. That may cause problems. I think I see what you mean: ctrl_id will be set to 0 if mshc alias is missing, and then caps is set as mmc->caps |= drv_data->caps[ctrl_id]; where drv_data->caps[0] = MMC_CAP_1_8V_DDR | MMC_CAP_8_BIT_DATA Should be fixed in next version of patchset then in any case, maybe it deserves a "Fixes:" tag as well, will have a look in git history if there is a particular commit that introduced this situation. > Best regards, > David Best regards, Henrik Grimler